Share your thoughts, 1 month free Claude Pro on usSee more

Science Question Answering on GPQA (accuracy)

63.51Accuracy

SFT

Updated 1mo ago

Evaluation Results

Method	Links
SFT 2026.04		63.51
D-QRELO 2026.04		61.47
Delta-CoMe 2026.04		58.46
SFT 2026.04		56.94
Llama-Instruct 2026.03		54.5
D-QRELO 2026.04		53.27
Delta-CoMe 2026.04		51.92
SVD 2026.04		50.43
R1-Distill-Qwen-7B 2026.03		49
Qwen-7B 2026.03		48
Qwen-7B 2026.03		48
Qwen-7B 2026.03		47
Llama 3.1 Instruct 2025.04		46.7
SFT 2026.04		45.2
Qwen-7B 2026.03		45
Wanda 2026.04		44.52
SVD 2026.04		44.34
Qwen-7B 2026.03		43
Magnitude 2026.04		42.74
ParamΔ 2025.04		42.2
Llama 3 Instruct 2025.04		40.6
Qwen-3B 2026.03		40
BitDelta 2026.04		39.43
Wanda 2026.04		39.04
Evo 8B 2026.02		38.4
Magnitude 2026.04		37.82
Qwen-3B 2026.03		37
Qwen-7B 2026.03		37
Backbone 2026.04		36.36
Qwen2.5 7B 2026.02		36.1
HATified-SFT 2026.03		36
D-QRELO 2026.04		35.98
BitDelta 2026.04		35.35
M_RLP + Post 2025.09		34.97
Random 2026.04		34.93
M_RLP + Post 2025.09		33.33
Llama 3 Instruct 2025.04		32.8
Backbone 2026.04		32.32
Qwen-3B 2026.03		32
M_base + Post 2025.09		31.52
Magnitude 2026.04		31.19
Qwen-3B 2026.03		31
M_base + Post 2025.09		30.93
Wanda 2026.04		30.72
ParamΔ 2025.04		30.6
BitDelta 2026.04		30.56
BD3-LM 7B 2026.02		30.5
Random 2026.04		30.39
SVD 2026.04		30.32
Delta-CoMe 2026.04		30.18
M_CPT + Post 2025.09		30.01
Llama 3.1 Instruct 2025.04		29.7
M_CPT + Post 2025.09		29.27
M_RLP 2025.09		28.28
M_base 2025.09		27.52
M_RLP 2025.09		27.02
LLaDA 8B 2026.02		26.4
M_CPT 2025.09		26.26
LLaMA3 8B 2026.02		26.2
ARD 7B 2026.02		25.3
M_base 2025.09		25.25
MDLM 7B 2026.02		24.9
M_CPT 2025.09		24.75
Llama 3.1 Base 2025.04		22.8
Random 2026.04		22.22
Backbone 2026.04		20.13
Llama 3 Base 2025.04		14.3
Llama 3.1 Base 2025.04		6.3
Llama 3 Base 2025.04		5.1